Functions

First we need to make some functions taking in a dataset, a variable in the dataset and a grouping variable, and returning a correctly formatted character string. I make two functions as an example, one returning the mean and standard deviation, and one a count with percentage. I also make one just returning empty strings for headers. Note the use of quasiquotation functions (ensym and !!). I am not very proefficient with such functions, but this works somehow.

mean_sd <- function(data, var, group, digits = 1) {
  var <- ensym(var)
  group <- ensym(group)
  data %>% 
    group_by(!!group) %>% 
    summarise(mean = mean(!!var, na.rm = TRUE), sd = sd(!!var, na.rm = TRUE), missing = sum(is.na(!!var))) %>% 
    mutate_at(vars(mean, sd), ~round(., digits = digits)) %>% 
    mutate(txt = paste0(mean, " (", sd, ")")) %>% 
    select(group, txt) %>% 
    deframe
}


n_pct <-  function(data, var, group, level = 1) {
  var <- ensym(var)
  group <- ensym(group)
  data %>% 
    filter(!is.na(!!var)) %>% 
    group_by(!!group, !!var) %>% 
    summarise(n = n()) %>% 
    group_by(!!group) %>% 
    mutate(tot = sum(n),
           pct = round(n/tot*100,digits = 1)) %>% 
    mutate(txt = paste0(n, " (", pct, "%)")) %>% 
    filter(!!var == !!level) %>% 
    ungroup %>% 
    select(group, txt) %>% 
    deframe
}

empty <- function(data, var, group, ...){
  group <- ensym(group)
  data %>% 
    group_by(!!group) %>% 
    summarise(n = n()) %>% 
    mutate(txt = "") %>% 
    select(group,txt) %>% 
    deframe
}

Table style tibble

Then we need a tibble defining how the table should look like. The first column is the variable description, the next is the function to be used, then the variable name, and finally other arguments for the function call.

dm_table <- tribble(
  ~text, ~f, ~var, ~param,
  "**Age at enrollment (years)**", "mean_sd", "age", list(digits = 1), 
  "**Symptom duration, n (%)**", "empty", "", list(),
  "Less than 1 day", "n_pct", "sympdurcat", list(level = "<1"),
  " 1-2 days", "n_pct", "sympdurcat", list(level = "1-2"),
  " 2-7 days", "n_pct", "sympdurcat", list(level = ">2-7"),
  "More than 7 days", "n_pct", "sympdurcat", list(level = ">7"),
  "**Signs and symptoms**", "empty", "", list(),
  "Dysuria, n(%)", "n_pct", "dysblcat", list(level = "Yes"),
  "Urinary urgency, n(%)", "n_pct", "urgblcat", list(level = "Abnormal"), 
  "Urinary frequency, n(%)", "n_pct", "polblcat", list(level = "Abnormal"), 
  "Visible hematuria, n(%)", "n_pct", "hemblcat", list(level = "Yes"), 
  "Symptom severity sum score*", "mean_sd", "symptotscore", list(digits = 2),  
  "Dysuria symptom score", "mean_sd", "dysbl", list(digits = 2), 
  "Urinary urgency symptom score", "mean_sd", "urgbl", list(digits = 2), 
  "Urinary frequency symptom score", "mean_sd", "polbl", list(digits = 2)
)

We need a wrapper to the rlang::exec-function to forward the other arguments to the function call:

stats_exec <- function(f, data, var, group, ...){
    exec(f, data, var, group, !!!(...))
}

Add data

Then we need some data. I was involved in a trial on the use of antibiotics in women with urinary tract infection (UTI), the IMUTI trial. In this trial the raw data are included in the publication for transparency. The baseline data in .rds format is stored here

imutibl <- readRDS("static/datasets/imutibl.rds") %>% 
  filter(fas == "Yes")

Purrr it!

Then everything is ready to use purrr to make the demographics and baseline characteristics table. I add the data as a list-column, and then use pmap to map the data, variable, grouping variable and arguments to the function using the stats_exec function. Then some unnesting, pivoting and kabelinig, and we have a nice table!

dm_table <- dm_table %>% 
  mutate(data = list(imutibl),
         group = "group") %>% 
  mutate(res = pmap(list(f, data, var, group, param), stats_exec)) %>% 
  mutate(id = map(res,names)) %>% 
  unnest(c(res, id)) %>% 
  pivot_wider(values_from = res, names_from = id) %>% 
  select(text, levels(imutibl$group)) 

knitr::kable(dm_table)

Then we can add some flourish, e.g. the number in each group in the header and footnotes:

header <- imutibl %>% 
  group_by(group) %>% 
  summarise(n=n()) %>% 
  mutate(txt = paste0(group, "\n (N=", n, ")")) %>% 
  select(txt) %>% 
  deframe
  
header <- c("Characteristic", header)

dm_table %>% knitr::kable(col.names = header) %>% 
  footnote(symbol = "Sum of day 0 symptopm scores of dysuria, urinary urgency and urinary frequency, range 0-18. The symptoms were given a value on a scale from 0 to 6 where 0 was 'normal/not affected'  and 6 was 'as bad as it could be'.",
           general = "SD, standard deviation",
           footnote_as_chunk = T)

Note that this table looks almost identical to Table 1 in the article! The magic of purrr…